Regulation of gene expression by the APP family in the adult cerebral cortex

Amyloid precursor protein (APP) is associated with both familial and sporadic forms of Alzheimer’s disease. APP has two homologs, amyloid precursor-like protein 1 and 2 (APLP1 and APLP2), and they have functional redundancy. APP intracellular c-terminal domain (AICD), produced by sequential α- or β- and γ-secretase cleavages, is thought to control gene expression, similarly as the ICD of Notch. To investigate the role of APP family in transcriptional regulation, we examined gene expression changes in the cerebral cortex of APP/APLP1/APLP2 conditional triple knockout (cTKO) mice, in which APP family members are selectively inactivated in excitatory neurons of the postnatal forebrain. Of the 12 previously reported AICD target genes, only Nep and Npas4 mRNA levels were significantly reduced in the cerebral cortex of cTKO mice, compared to littermate controls. We further examined global transcriptional changes by RNA-seq and identified 189 and 274 differentially expressed genes in the neocortex and hippocampus, respectively, of cTKO mice relative to controls. Gene Ontology analysis indicated that these genes are involved in a variety of cellular functions, including extracellular organization, learning and memory, and ion channels. Thus, inactivation of APP family alters transcriptional profiles of the cerebral cortex and affects wide-ranging molecular pathways.

www.nature.com/scientificreports/ To provide molecular insight into the physiological and signaling function of APP family members in the adult brain, we used recently generated APP/APLP1/APLP2 triple conditional knockout (cTKO) mice 23 , in which all APP family members are selectively inactivated in excitatory neurons postnatally in the forebrain. We evaluated the expression of previously reported putative downstream target genes of AICD or ALIDs in the neocortex and hippocampus at the age of 3 months. In addition, unbiased analysis of global gene expression using RNA-seq revealed that loss of the APP family alters the transcription profile in the adult mouse cerebral cortex. These findings represent an initial step toward a more complete understanding of the role of APP family genes in the adult brain.

Results
Downregulation of the expression levels of Nep and Npas4 in the cerebral cortex of cTKO mice. Our first goal was to determine whether the expression levels of previously identified putative AICD downstream target genes are changed in the absence of APP, APLP1, and APLP2 in the adult brain. Due to the genetic redundancy of APP, APLP1 and APLP2 and the perinatal lethality of APP/APLP1/APLP2 germline triple knockout (TKO) mice, it is not possible to study the function of the APP family in the adult brain 24 . Thus, we used APP/APLP1/APLP2 conditional TKO (fAPP/fAPP; fAPLP1/fAPLP2; fAPLP2/fAPLP2; Camk2a-Cre; hereafter cTKO) mice, in which all APP family members are selectively inactivated in excitatory neurons postnatally in the forebrain 23 . These cTKO mice show mild learning and memory deficits as well as impairment of synaptic plasticity at the age of 3 months, but there is no gross abnormality in the cerebral cortex of cTKO up to ~ 2 years of age. We performed qRT-PCR of total RNA extracts from the neocortex and hippocampus of cTKO and control mice (fAPP/fAPP; fAPLP1/fAPLP2; fAPLP2/fAPLP2) at age 3 months. Consistent with previous Northern analysis results 23 , we observed a significant decrease in the mRNA expression of App, Aplp1, and Aplp2 in the neocortex (App: 76%, p < 0.0001; Aplp1: 73%, p < 0.0001, Aplp2: 67%, p < 0.0001; Fig. 1a) and hippocampus (App: 71%, p < 0.0001; Aplp1: 71%, p < 0.0001; Aplp2: 59%, p < 0.0001, Student's t-test, n = 5 per genotype; Fig. 1b) of cTKO compared to littermate control mice. Figure 1. Expression levels of APP family genes and putative AICD target genes in the cerebral cortex of cTKO mice at 3 months of age. (a,b) Relative mRNA levels of putative AICD target genes in the neocortex (a) and hippocampus (b) of cTKO mice relative to the controls at 3 months of age. The mRNA levels of each gene were normalized to the mRNA levels of the Gapdh and Rpl13a housekeeping genes. All data are expressed as the means ± SEMs. n = 5-6 mice per genotype. Statistical analysis was performed with Student's t-test. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. www.nature.com/scientificreports/ Then, we evaluated the expression levels of previously reported putative AICD downstream target genes (Kai, Gsk3b, Bace1, Tip60, Nep, P53, Egfr, Lrp1, Fe65, vGlut2, X11, and Npas4) 12,[14][15][16][17][18][19][20][21][22] in the cerebral cortex of cTKO and control mice by qRT-PCR analysis. We detected a significant decrease in the levels of Nep (1st primer set: 18%, p = 0.0296; 2nd primer set: 34%, p = 0.0046, Student's t-test) and Npas4 (1st primer set: 41%, p = 0.0086; 2nd primer set: 43%, p = 0.0053, Student's t-test) in the neocortical transcriptome in cTKO mice compared to controls (Fig. 1a). Similarly, the expression levels of Nep and Npas4 in the hippocampus were significantly decreased in cTKO mice compared to controls (Nep; 1st primer set: 21%, p = 0.0009; 2nd primer set: 21%, p = 0.0076, Npas4; 1st primer set: 60%, p = 0.0019; 2nd primer set: 60%, p = 0.0018, Student's t-test, Fig. 1b). However, the expression levels of other putative AICD downstream target genes showed no significant difference between cTKO and control mice. Together, loss of all APP family genes in the cerebral cortex in vivo reduces gene expression of Nep and Npas4, which have been reported as AICD target genes in other animal model systems.

Scientific Reports
Differential gene expression in the cerebral cortex of mice lacking APP family gene in excitatory neurons. Having confirmed that loss of APP family genes can alter gene expression of Nep and Npas4 in the cerebral cortex, we then aimed to assess the global impact of APP family gene loss on transcriptional changes by performing RNA-seq experiments using the neocortex and hippocampus of cTKO and littermate control mice (n = 5 per genotype) at the age of 3 months. A total of 86-91% of raw reads were mapped to the mm10 mouse genome (Supplementary Table 1). Multi-dimensional scaling (MDS) plots indicated that gene expression profiles were more distinct between different tissues than genetic changes of between cTKO mice and controls (Fig. 2a-c).
We then compared the gene expression in the neocortex and hippocampus of cTKO mice and littermate controls using edgeR 25 Table S2). We then performed technical validation of our RNA-seq dataset by qRT-PCR of cerebral cortex obtained from a different set of mice. First, we confirmed that App, Aplp1, and Aplp2 expression was, as expected, significantly lower in cTKO mice than in the controls (Fig. 3a,b). We performed qRT-PCR for the set of DEGs that were differentially expressed in both the neocortex and hippocampus (42 genes). The qRT-PCR results showed that of these 42 genes, 39 and 41 showed identical patterns to those determined by RNA-seq in the neocortex and hippocampus, respectively. However, for 3 genes (Fndc9, Foxm1, Top2a), although qRT-PCR confirmed the direction of the changes in mean expression (higher or lower) in cTKO mice, the difference was not statistically significant. Thus, our validation rate was 93% for the neocortex and 98% for the hippocampus. This finding suggests that the expressions of these genes are affected by APP family signaling universally with spatial restriction. Overall, these data demonstrate that the inactivation of the APP family in the cerebral cortex alters the transcription profile of a large set of genes with diverse roles.

Changes of molecular pathways in the cerebral cortex by the loss of APP family.
To explore the biological processes and molecular functions of the genes that were differentially regulated in the cerebral cortex of cTKO mice compared to that of controls, we performed Gene Ontology (GO) enrichment analyses using the Goseq R package 26 . The GO terms enriched among the DEGs with FDR-corrected p-values of less than 0.05 were identified. The DEGs in the neocortex and hippocampus were annotated with 86 and 133 GO biological process terms, respectively. In particular, the extracellular structure organization, calcium ion regulated exocytosis, and inhibitory postsynaptic potential were top enriched pathways in biological processes affected by the loss of APP family in the neocortex (Fig. 4a). Interestingly, genes associated with the neuropeptide signaling pathway, regulation of membrane potential, and learning-or memory-related biological processes were most enriched pathways in the hippocampus (Fig. 4b). In parallel, a total of 20 (neocortex) and 53 (hippocampus) GO molecular function terms were identified with an FDR-corrected p-value of less than 0.05. Consistent with the biological processes GO terms, extracellular matrix structural constituents and channel activity were observed as most changed in the neocortex and hippocampus, respectively (Fig. 4c,d).
We also performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis 27 to investigate whether the DEGs associated with specific biochemical pathways. Among the 189 DEGs in the neocortex, significant enrichment was found in five KEGG pathways (adjusted p-value < 0.05). The most significantly overrepresented pathway was the AGE-RAGE signaling pathway in diabetic complications. Similarly, significant enrichment of protein digestion and absorption and the MAPK signaling pathway was observed (Fig. 4e)  www.nature.com/scientificreports/ the hippocampus DEGs, nine KEGG pathways (adjusted p-value < 0.05) were significantly enriched. Interestingly, neuroactive ligand-receptor interactions and calcium signaling pathways were significantly overrepresented (Fig. 4f). Overall, our GO analysis indicates that loss of the APP family genes in excitatory neurons of the cerebral cortex has very significant impacts on the extracellular structure, ion channel activity, and neuropeptide signaling processes.

Discussion
Transcriptome analysis of the postmortem human brain tissue [28][29][30][31][32] or genetically modified animal models [33][34][35][36][37][38] has provided a comprehensive overview of the transcriptional responses associated with AD pathogenesis. In particular, AICD has been proposed as a key epigenetic regulator of gene expression controlling a diverse range of genes. However, the validity of its proposed targets, and particularly regarding the question of whether they also constitute endogenous AICD targets in the adult brain, has remained controversial due to the presence of functional compensation of two other APP homologs, APLP1 and APLP2, and the lack of adequate animal models 39 . The creation of a new animal model, in which all APP family genes are selectively inactivated in excitatory neurons in the adult mouse brain, allowed us to investigate the changes in gene expression in the adult cerebral cortex, and the functions of APP family genes at the global level during their normal function. The qRT-PCR analysis revealed that among the proposed AICD interacting target genes, only Nep and Npas4 exhibited significantly changed expression in the cerebral cortex of cTKO mice compared to controls. APP dependent expression changes of Nep or Npas4 were previously investigated in the primary cortical neurons, fibroblasts or brains of APP knockout mice 18,22 , demonstrating that our findings are highly consistent with the results of previous studies using in vivo models. Conversely, the expression levels of other proposed target genes, which were previously identified and tested using in vitro culture systems, were similar in cTKO mice and controls. www.nature.com/scientificreports/ Our RNA-seq analysis provides a comprehensive understating of the changes in transcriptional regulation that occur in the cerebral cortex in the absence of the APP family in excitatory neurons. Based on the 189 and 274 DEGs found in the neocortex and hippocampus, respectively, we performed an enrichment analysis using the KEGG and GO databases. The list of pathways obtained from this analysis may help elucidate the molecular mechanisms of phenotypes found in knockout models of the APP family. For example, we previously reported that the APP family regulates the synaptic plasticity and neuronal excitability through Kv7 channels in the hippocampal Schaffer Collateral synapses and CA1 neurons 23 . Consistent with these results, we found that membrane potential-and ion channel activity-related gene sets, particularly voltage-gated potassium channelencoding genes (e.g. Kcna5, Kcne1l, Kcnh5, and Kcnq3 etc.), were enriched in the hippocampus. Interestingly, the mRNA levels of Kcnq3, one of Kv7 encoding genes, were significantly decreased in the hippocampal transcriptome of cTKO mice compared to controls. The qRT-PCR analysis confirmed that only the expression levels of Kcnq3, but not other Kv7 channel encoding genes, were significantly reduced in the hippocampus of cTKO mice relative to controls (Supplementary Fig. S1). The molecular pathway through which the APP family regulates Kv7 channel transcription and activity during the regulation of neuronal excitability is worth further investigation, and its elucidation may be further explored to identify novel therapeutic targets in AD.
APP family genes are highly expressed in excitatory neurons of the cerebral cortex but are not exclusive to these cells, as their gene products persist at some level in excitatory neuron-specific cTKO mice 23 . APP is also expressed in other types of cells, such as GABAergic interneurons and glial cells 40,41 . In addition, various protein fragments derived from APP regulate synaptic plasticity and neural networks [42][43][44][45][46] . Thus, we cannot exclude that www.nature.com/scientificreports/ cell-nonautonomous effects of the loss of APP family function in excitatory neurons on other types of cells have been observed. For example, the expression levels of Npy, which is secreted from GABAergic interneurons 47 , were significantly increased in the hippocampal transcriptome of cTKO mice. Hence, it would be interesting to investigate the cell type specific transcriptional changes that occur in the absence of the APP family at higher resolution by single-cell RNA-seq or by generating and analyzing inhibitory neuron-or glia-specific cTKO mice.
In conclusion, excitatory neuron-specific deletion of the APP family in the adult brain leads to changes in gene expression involved in a wide range of molecular pathways. The presented model can be used in future studies to investigate the interactions of the APP family with voltage-gated potassium channels as well as influence of APP family genes on other types of cells. Our work provides the first description of the transcriptome in cTKO mice. We expect this will provide a starting point for better understanding the function of APP family genes in the cerebral cortex.

Isolation of the neocortex and hippocampus.
Mice were euthanized using the controlled release of carbon dioxide followed by cervical dislocation, and the brains were removed. The neocortex and hippocampus were dissected separately from the sagittal cut brain on ice and stored at − 80 °C until further processing.
Total RNA extraction and qRT-PCR. Total RNA was extracted from the neocortex and hippocampus using the RNeasy Plus Universal Kit and RNeasy Plus Mini kit (Qiagen), respectively, according to the manufacturer's instructions. Approximately 1 µg of RNA was reverse transcribed using iScript Reverse Transcription Supermix (Bio-Rad) according to the manufacturer's instructions. For qPCR, each 10 µl reaction contained 5 µl of PowerUP SYBR Green PCR Master Mix (Thermo Fisher Scientific), 100 nM of each primer, and template cDNA corresponding to 250 ng of RNA. cDNA was denatured at 95 °C for 2 min and amplified for 45 cycles with a QuantStudio 6 (Thermo Fisher Scientific) using a standard protocol. The sequences of the primers can be found in Supplementary Table S3. The specificity of the primers was determined by the single peak in the melt curve. All reactions were performed in at least duplicates. The gene expression levels were normalized to the expression levels of Gapdh and Rpl13a.
RNA-seq analysis. The neocortex and hippocampus were collected separately from five cTKO and five littermate controls at the age of 3 months, and total RNA was extracted as described above. The RNA quality of 20 samples was analyzed with a 2100 BioAnalyzer (Agilent Technologies) using the RNA 6000 Nano kit (Agilent Technologies), and all 20 samples had an RNA integrity number (RIN) between 8.1 and 9.8 (Supplementary  Table S1) 49 . To enrich mRNAs from the total RNA sample (200 ng-1 µg per sample), poly-A pull-down was performed, and library preparation was conducted using an Illumina TruSeq RNA prep kit (Illumina). The libraries were then sequenced using Illumina HiSeq2500/HiSeq4000 with 100-bp single-end reads per sample at Columbia Genome Center (https:// syste msbio logy. colum bia. edu/ genome-center). The entire RNA-seq dataset is available at the Gene Expression Omnibus (GEO) database with the accession number GSE181936.
The sequencing reads were mapped to the mm10 mouse reference genome using STAR (v2.5.4a) 50 with the default parameters. Raw counts of the aligned reads were produced using HTseq 51 . Differential expression analyses were then performed as described 52 using edgeR 25 to identify significantly differentially expressed genes in the neocortex and hippocampus tissues between cTKO mice and controls. After the trimmed mean of M-values (TMM) normalization, multi-dimensional scaling (MDS) plots were generated based on leading log-fold-change distances and the first two dimensions were selected to illustrate the data. Enrichment analysis of gene ontology (GO) terms. DEGs (FDR < 0.05) in cTKO mice compared to the control mice were selected for enrichment analyses of gene sets. The Goseq R package 26 was used to determine the enrichment of GO terms and KEGG pathways taking into account gene length bias. Adjusted p-values were calculated using the Benjamini-Hochberg method, and GO gene sets with p-values less than 0.05 were considered significant. In the KEGG pathway enrichment analysis, gene sets with p-values less than 0.05 were considered significant.

Data availability
The entire RNA-seq dataset is available at the Gene Expression Omnibus (GEO) database with the accession number GSE181936. Other datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.